TEST RESULTS AND DISCUSSION Sample Clauses

TEST RESULTS AND DISCUSSION. The algorithm that directly applies single-loop measurements to Equation (2-1) for speed estimates by using a constant g is identified as the “traditional algorithm” in this report. The traditional algorithm and the proposed region growing algorithm were tested against ground truth data for periods of 9, 12, and 15 20-second intervals (3, 4, and 5 minutes, respectively). Graphs of the actual versus plotted values with R2 values are provided in Figure 4-1, and a summary of the results is provided in Table 4-2. A perfect estimation would result in all data points forming a line of slope 1.0 starting at the origin. Therefore, data points falling under the ideal line were underestimated speeds, and those above the line were overestimated speeds. The proposed algorithm, based on the revised region growing concept, clearly provided superior speed estimates. The results of LV estimation are provided in Table 4-3. Comparisons are given in absolute differences for the entire day. Computation of more complex error measurements did not seem appropriate because LV volumes in general constituted less than 10 percent of the traffic at each location and, therefore, could be considered a somewhat “rare” event. Daily LV volume estimates were on average within 4.0 percent of the dual-loop estimated LV volumes. Station & Loop Code Loop Coeff. Beta Period Length (min) SSE SSE / Period Average % Error SSE SSE / Period Average % Error SSE SSE / Period Average % Error ES-167D _MS 2 1.01 345 38504 24339 16471 80 68 57 11.6% 10.7% 10.1% 12593 6369 4124 26 18 14 6.3% 5.5% 5.1% ▇▇▇▇▇ ▇▇▇▇ ▇▇▇▇ ▇▇ 21 16 6.2% 5.9% 5.3% ES-172R MMS 2 0.92 345 34421 21293 14681 149 59 144 11.1% 10.1% 9.4% 16698 7208 4209 36 20 15 7.2% 6.1% 5.4% 10109 6275 3735 19 17 11 6.2% 5.7% 5.0% ES-209D _MN 2 1.01 345 34650 21713 15815 255 60 257 11.3% 10.4% 9.8% 15265 9800 6550 34 27 25 6.8% 6.5% 6.0% 13421 8808 5309 25 24 17 6.2% 5.8% 5.3% Station & Loop Code Period Length (min) Dual-Loop LV Volume Estimated LV Volume Error % Error ES-167D _MS 2 3 4 5 ▇▇▇▇ ▇▇▇▇ ▇▇▇▇ 2315 2317 2285 -54 -52 -84 2.28% 2.20% 3.55% ES-172R MMS 2 3 4 5 2566 2566 2566 ▇▇▇▇ ▇▇▇▇ ▇▇▇▇ 112 117 156 4.36% 4.56% 6.08% ES-209D _MN 2 3 4 5 2630 2630 2630 2823 2760 2602 193 130 -28 7.34% 4.94% 1.06% Because it is true that intervals containing LVs would naturally have a higher occupancy variance, one might question, on the basis of heteroskedasticity concerns, whether it is valid to use R2 values as a measure of goodness-of-fit for speed estimat...